Improving Neural Text Summarization using Knowledge Graphs

Abstract

In this paper, we propose a method for extractive text summarization using auto-regressive transformers. For better learning procedure we adopt the knowledge graph method to convert our textual data to more informative text and unsupervised training methods for wide use. We feed the informative text to our pre-trained generative model to summarize the text more properly and infer on generating a proper summary. The model is able to summarize input text into adequate information and is capable of performing several Natural Language Processing(NLP) tasks.

Tasks

Reproductions